We asked 5 AI helpers to write tough emails. One was a clear winner.

We asked 5 AI helpers to write tough emails. One was a clear winner.

TL;DR:

• A recent evaluation of AI email writing assistants tested five tools: ChatGPT, Claude, Copilot, DeepSeek, and Gemini.
• Anthropic's Claude emerged as the best performer, surpassing not just its AI counterparts but a human writer as well.
• Each AI was asked to draft emails ranging from apologies to complex announcements, with an expert panel scoring their outputs.
• The authenticity and emotional resonance of AI-generated emails significantly impacted their effectiveness.

In an era where communication is pivotal for personal and professional interactions, the application of artificial intelligence (AI) in drafting emails has become increasingly commonplace. To determine which AI email writing assistant is the most effective, The Washington Post conducted a unique evaluation involving five leading AI tools: ChatGPT, Claude, Copilot, DeepSeek, and Gemini. A panel of communications experts rigorously tested these tools, ultimately declaring Claude from Anthropic the standout winner in their writing capabilities.

Methodology of the Evaluation

The evaluation consisted of an "old-fashioned bake-off," where each AI was tasked with drafting five challenging types of emails. The prompts included:

1. Apology letter to a friend.
2. Layoff announcement from a CEO to employees.
3. A humorous request urging a spouse to move to the North Pole.
4. A proposal to convert an office break room into a ball pit.
5. A delicate breakup text.

After drafting the emails, a blue-ribbon panel ranked the submissions, evaluating them on criteria such as clarity, warmth, and emotional accuracy[^1].

Results of the Evaluation

The results from the judges were telling. Here’s how the AIs ranked in effectiveness from least to most proficient:

1. Microsoft Copilot – Score: 23/100

• The judges labeled Copilot’s writing as "robotic" with generic phrases, failing to resonate emotionally[^2].

1. OpenAI’s ChatGPT – Score: 43/100

• While its emails were direct, they often came across as "transactional" and stilted.

1. Google’s Gemini – Score: 44/100

• Gemini's emails had clarity but lacked human warmth, often sounding distinctly artificial[^3].

1. DeepSeek – Score: 45/100

• Although better than the earlier entries, DeepSeek’s emails were criticized for verbosity and overly complex wording.

1. Anthropic’s Claude – Score: 50/100

• Claude excelled by communicating genuine human emotion and understanding, even outperforming the human writer tested alongside the AIs[^4].

Implications of the Findings

The panel’s feedback highlighted significant trends regarding AI writing technologies. Despite the technical proficiency of these tools, the judges emphasized a core issue—authenticity. Successful email communication requires not just correctness but also true human connection and emotional context. Claude’s success stems from its ability to create messages that felt genuine and relatable.

This evaluation spotlights the ongoing evolution and competitive landscape of AI communication tools. As these technologies become integrated into platforms like Gmail and Outlook, understanding their capabilities—and limitations—will be critical for users seeking efficient yet heartfelt communication[^5].

Conclusion

The study underscores the growing role of AI in personal and professional communication, marking the importance of choosing the right tool for drafting sensitive messages. As Anthropic’s Claude sets a new benchmark for AI writing, users should consider both effectiveness and emotional resonance when selecting AI tools for email composition.

The implications of AI on communication are profound, suggesting a future where technology not only assists but potentially enhances human connection in digital correspondence.

References

[^1]: The Washington Post (Mar 26, 2025). "We asked 5 AI helpers to write tough emails. One was a clear winner.". Retrieved March 28, 2025.

[^2]: Geoffrey Fowler (Mar 26, 2025). "Review | We asked 5 AI helpers to write tough emails. One was a clear winner.". The Washington Post. Retrieved March 28, 2025.

[^3]: The Washington Post (Mar 26, 2025). "The best AI email writing assistant: We tested 5, and only one beats a human". Retrieved March 28, 2025.

[^4]: The Washington Post. "What did we learn?". Retrieved March 28, 2025.

[^5]: Geoffrey Fowler (Mar 26, 2025). "Full results and samples from my AI writing test at this @washingtonpost gift link:". Retrieved March 28, 2025.

Keywords: AI email assistants, Anthropic Claude, ChatGPT, Microsoft Copilot, DeepSeek, Gemini, AI communication, email writing evaluation.